This obviates the patch that teaches LLVM internals about
_rust_{re,de}alloc functions by putting annotations directly in the IR
for the optimizer.
The sole test change is required to anchor FileCheck to the body of the
`box_uninitialized` method, so it doesn't see the `allocalign` on
`__rust_alloc` and get mad about the string `alloca` showing up. Since I
was there anyway, I added some checks on the attributes to prove the
right attributes got set.
While we're here, we also emit allocator attributes on
__rust_alloc_zeroed. This should allow LLVM to perform more
optimizations for zeroed blocks, and probably fixes#90032. [This
comment](https://github.com/rust-lang/rust/issues/24194#issuecomment-308791157)
mentions "weird UB-like behaviour with bitvec iterators in
rustc_data_structures" so we may need to back this change out if things
go wrong.
The new test cases require LLVM 15, so we copy them into LLVM
14-supporting versions, which we can delete when we drop LLVM 14.
Add support for LLVM ShadowCallStack.
LLVMs ShadowCallStack provides backward edge control flow integrity protection by using a separate shadow stack to store and retrieve a function's return address.
LLVM currently only supports this for AArch64 targets. The x18 register is used to hold the pointer to the shadow stack, and therefore this only works on ABIs which reserve x18. Further details are available in the [LLVM ShadowCallStack](https://clang.llvm.org/docs/ShadowCallStack.html) docs.
# Usage
`-Zsanitizer=shadow-call-stack`
# Comments/Caveats
* Currently only enabled for the aarch64-linux-android target
* Requires the platform to define a runtime to initialize the shadow stack, see the [LLVM docs](https://clang.llvm.org/docs/ShadowCallStack.html) for more detail.
Allow to disable thinLTO buffer to support lto-embed-bitcode lld feature
Hello
This change is to fix issue (https://github.com/rust-lang/rust/issues/84395) in which passing "-lto-embed-bitcode=optimized" to lld when linking rust code via linker-plugin-lto doesn't produce the expected result.
Instead of emitting a single unified module into a llvmbc section of the linked elf, it emits multiple submodules.
This is caused because rustc emits the BC modules after running llvm `createWriteThinLTOBitcodePass` pass.
Which in turn triggers a thinLTO linkage and causes the said issue.
This patch allows via compiler flag (-Cemit-thin-lto=<bool>) to select between running `createWriteThinLTOBitcodePass` and `createBitcodeWriterPass`.
Note this pattern of selecting between those 2 passes is common inside of LLVM code.
The default is to match the old behavior.
Remove branch target prologues from `#[naked] fn`
This patch hacks around rust-lang/rust#98768 for now via injecting appropriate attributes into the LLVMIR we emit for naked functions. I intend to pursue this upstream so that these attributes can be removed in general, but it's slow going wading through C++ for me.
Revert "Work around invalid DWARF bugs for fat LTO"
Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See #90357 for more info
and a repro case.
This commit is a mechanical revert of
d5de680e20 from PR #89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)
I have locally verified that this fixes#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
Adding the option to control from rustc CLI
if the resulted ".o" bitcode module files are with
thinLTO info or regular LTO info.
Allows using "-lto-embed-bitcode=optimized" during linkage
correctly.
Signed-off-by: Ziv Dunkelman <ziv.dunkelman@nextsilicon.com>
This adds the typeid and `vcall_visibility` metadata to vtables when the
-Cvirtual-function-elimination flag is set.
The typeid is generated in the same way as for the
`llvm.type.checked.load` intrinsic from the trait_ref.
The offset that is added to the typeid is always 0. This is because LLVM
assumes that vtables are constructed according to the definition in the
Itanium ABI. This includes an "address point" of the vtable. In C++ this
is the offset in the vtable where information for RTTI is placed. Since
there is no RTTI information in Rust's vtables, this "address point" is
always 0. This "address point" in combination with the offset passed to
the `llvm.type.checked.load` intrinsic determines the final function
that should be loaded from the vtable in the
`WholeProgramDevirtualization` pass in LLVM. That's why the
`llvm.type.checked.load` intrinsics are generated with the typeid of the
trait, rather than with that of the function that is called. This
matches what `clang` does for C++.
The vcall_visibility metadata depends on three factors:
1. LTO level: Currently this is always fat LTO, because LLVM only
supports this optimization with fat LTO.
2. Visibility of the trait: If the trait is publicly visible, VFE
can only act on its vtables after linking.
3. Number of CGUs: if there is more than one CGU, also vtables with
restricted visibility could be seen outside of the CGU, so VFE can
only act on them after linking.
To reflect this, there are three visibility levels: Public, LinkageUnit,
and TranslationUnit.
To apply the optimization the `Virtual Function Elim` module flag has to
be set. To apply this optimization post-link the `LTOPostLink` module
flag has to be set.
In https://reviews.llvm.org/D125556 upstream changed sext() and zext()
to allow some no-op cases, which previously required use of the *OrSelf()
methods, which I assume is what was going on here. The *OrSelf() methods
got removed in https://reviews.llvm.org/D125559 after two weeks of
deprecation because they came with some bonus (probably-undesired)
behavior. Since the behavior of sext() and zext() changed slightly, I
kept the old *OrSelf() calls in LLVM 14 and earlier, and only use the
new version in LLVM 15.
r? @nikic
This new enum entry was introduced in https://reviews.llvm.org/D122268,
and if I'm reading correctly there's no case where we'd ever encounter
it in our uses of LLVM. To preserve the ability to compile this file
with -Werror -Wswitch we add an explicit case for this entry.
Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See #90357 for more info
and a repro case.
This commit is a mechanical revert of
d5de680e20 from PR #89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)
I have locally verified that this fixes#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
Remove LLVM attribute removal
This was necessary before, because `declare_raw_fn` would always apply
the default optimization attributes to every declared function.
Then `attributes::from_fn_attrs` would have to remove the default
attributes in the case of, e.g. `#[optimize(speed)]` in a `-Os` build.
(see [`src/test/codegen/optimize-attr-1.rs`](03a8cc7df1/src/test/codegen/optimize-attr-1.rs (L33)))
However, every relevant callsite of `declare_raw_fn` (i.e. where we
actually generate code for the function, and not e.g. a call to an
intrinsic, where optimization attributes don't [?] matter)
calls `from_fn_attrs`, so we can remove the attribute setting
from `declare_raw_fn`, and rely on `from_fn_attrs` to apply the correct
attributes all at once.
r? `@ghost` (blocked on #94221)
`@rustbot` label S-blocked
This was necessary before, because `declare_raw_fn` would always apply
the default optimization attributes to every declared function,
and then `attributes::from_fn_attrs` would have to remove the default
attributes in the case of, e.g. `#[optimize(speed)]` in a `-Os` build.
However, every relevant callsite of `declare_raw_fn` (i.e. where we
actually generate code for the function, and not e.g. a call to an
intrinsic, where optimization attributes don't [?] matter)
calls `from_fn_attrs`, so we can simply remove the attribute setting
from `declare_raw_fn`, and rely on `from_fn_attrs` to apply the correct
attributes all at once.
Add MemTagSanitizer Support
Add support for the LLVM [MemTagSanitizer](https://llvm.org/docs/MemTagSanitizer.html).
On hardware which supports it (see caveats below), the MemTagSanitizer can catch bugs similar to AddressSanitizer and HardwareAddressSanitizer, but with lower overhead.
On a tag mismatch, a SIGSEGV is signaled with code SEGV_MTESERR / SEGV_MTEAERR.
# Usage
`-Zsanitizer=memtag -C target-feature="+mte"`
# Comments/Caveats
* MemTagSanitizer is only supported on AArch64 targets with hardware support
* Requires `-C target-feature="+mte"`
* LLVM MemTagSanitizer currently only performs stack tagging.
# TODO
* Tests
* Example
In https://reviews.llvm.org/D114543 the uwtable attribute gained a flag
so that we can ask for sync uwtables instead of async, as the former are
much cheaper. The default is async, so that's what I've done here, but I
left a TODO that we might be able to do better.
While in here I went ahead and dropped support for removing uwtable
attributes in rustc: we never did it, so I didn't write the extra C++
bridge code to make it work. Maybe I should have done the same thing
with the `sync|async` parameter but we'll see.
This doesn't handle `char` because it's a bit awkward to distinguish it
from u32 at this point in codegen.
Note that for some types (like `&Struct` and `&mut Struct`),
we already apply `dereferenceable`, which implies `noundef`,
so the IR does not change.
This agrees with Clang, and avoids an error when using LTO with mixed
C/Rust. LLVM considers different behaviour flags to be a mismatch,
even when the flag value itself is the same.
This also makes the flag setting explicit for all uses of
LLVMRustAddModuleFlag.
This was originally introduced in #10916 as a way to remove all landing
pads when performing LTO. However this is no longer necessary today
since rustc properly marks all functions and call-sites as nounwind
where appropriate.
In fact this is incorrect in the presence of `extern "C-unwind"` which
must create a landing pad when compiled with `-C panic=abort` so that
foreign exceptions are caught and properly turned into aborts.
RustWrapper: adapt to new AttributeMask API
Upstream LLVM change 9290ccc3c1a1 migrated attribute removal to use
AttributeMask instead of AttrBuilder, so we need to follow suit here.
r? ``@nagisa`` cc ``@nikic``
No functional changes intended.
The LLVM commit
ec501f15a8
removed the signed version of `createExpression`. This adapts the Rust
LLVM wrappers accordingly.
Mark drop calls in landing pads `cold` instead of `noinline`
Now that deferred inlining has been disabled in LLVM (#92110), this shouldn't cause catastrophic size blowup.
I confirmed that the test cases from https://github.com/rust-lang/rust/issues/41696#issuecomment-298696944 still compile quickly (<1s) after this change. ~Although note that I wasn't able to reproduce the original issue using a recent rustc/llvm with deferred inlining enabled, so those tests may no longer be representative. I was also unable to create a modified test case that reproduced the original issue.~ (edit: I reproduced it on CI by accident--the first commit timed out on the LLVM 12 builder, because I forgot to make it conditional on LLVM version)
r? `@nagisa`
cc `@arielb1` (this effectively reverts #42771 "mark calls in the unwind path as !noinline")
cc `@RalfJung` (fixes#46515)
edit: also fixes#87055
Add support for LLVM coverage mapping format versions 5 and 6
This PR cherry-pick's Swatinem's initial commit in unsubmitted PR #90047.
My additional commit augments Swatinem's great starting point, but adds full support for LLVM
Coverage Mapping Format version 6, conditionally, if compiling with LLVM 13.
Version 6 requires adding the compilation directory when file paths are
relative, and since Rustc coverage maps use relative paths, we should
add the expected compilation directory entry.
Note, however, that with the compilation directory, coverage reports
from `llvm-cov show` can now report file names (when the report includes
more than one file) with the full absolute path to the file.
This would be a problem for test results, but the workaround (for the
rust coverage tests) is to include an additional `llvm-cov show`
parameter: `--compilation-dir=.`
Emit LLVM optimization remarks when enabled with `-Cremark`
The default diagnostic handler considers all remarks to be disabled by
default unless configured otherwise through LLVM internal flags:
`-pass-remarks`, `-pass-remarks-missed`, and `-pass-remarks-analysis`.
This behaviour makes `-Cremark` ineffective on its own.
Fix this by configuring a custom diagnostic handler that enables
optimization remarks based on the value of `-Cremark` option. With
`-Cremark=all` enabling all remarks.
Fixes#90924.
r? `@nikic`
LLVM has built-in heuristics for adding stack canaries to functions. These
heuristics can be selected with LLVM function attributes. This patch adds a
rustc option `-Z stack-protector={none,basic,strong,all}` which controls the use
of these attributes. This gives rustc the same stack smash protection support as
clang offers through options `-fno-stack-protector`, `-fstack-protector`,
`-fstack-protector-strong`, and `-fstack-protector-all`. The protection this can
offer is demonstrated in test/ui/abi/stack-protector.rs. This fills a gap in the
current list of rustc exploit
mitigations (https://doc.rust-lang.org/rustc/exploit-mitigations.html),
originally discussed in #15179.
Stack smash protection adds runtime overhead and is therefore still off by
default, but now users have the option to trade performance for security as they
see fit. An example use case is adding Rust code in an existing C/C++ code base
compiled with stack smash protection. Without the ability to add stack smash
protection to the Rust code, the code base artifacts could be exploitable in
ways not possible if the code base remained pure C/C++.
Stack smash protection support is present in LLVM for almost all the current
tier 1/tier 2 targets: see
test/assembly/stack-protector/stack-protector-target-support.rs. The one
exception is nvptx64-nvidia-cuda. This patch follows clang's example, and adds a
warning message printed if stack smash protection is used with this target (see
test/ui/stack-protector/warn-stack-protector-unsupported.rs). Support for tier 3
targets has not been checked.
Since the heuristics are applied at the LLVM level, the heuristics are expected
to add stack smash protection to a fraction of functions comparable to C/C++.
Some experiments demonstrating how Rust code is affected by the different
heuristics can be found in
test/assembly/stack-protector/stack-protector-heuristics-effect.rs. There is
potential for better heuristics using Rust-specific safety information. For
example it might be reasonable to skip stack smash protection in functions which
transitively only use safe Rust code, or which uses only a subset of functions
the user declares safe (such as anything under `std.*`). Such alternative
heuristics could be added at a later point.
LLVM also offers a "safestack" sanitizer as an alternative way to guard against
stack smashing (see #26612). This could possibly also be included as a
stack-protection heuristic. An alternative is to add it as a sanitizer (#39699).
This is what clang does: safestack is exposed with option
`-fsanitize=safe-stack`.
The options are only supported by the LLVM backend, but as with other codegen
options it is visible in the main codegen option help menu. The heuristic names
"basic", "strong", and "all" are hopefully sufficiently generic to be usable in
other backends as well.
Reviewed-by: Nikita Popov <nikic@php.net>
Extra commits during review:
- [address-review] make the stack-protector option unstable
- [address-review] reduce detail level of stack-protector option help text
- [address-review] correct grammar in comment
- [address-review] use compiler flag to avoid merging functions in test
- [address-review] specify min LLVM version in fortanix stack-protector test
Only for Fortanix test, since this target specifically requests the
`--x86-experimental-lvi-inline-asm-hardening` flag.
- [address-review] specify required LLVM components in stack-protector tests
- move stack protector option enum closer to other similar option enums
- rustc_interface/tests: sort debug option list in tracking hash test
- add an explicit `none` stack-protector option
Revert "set LLVM requirements for all stack protector support test revisions"
This reverts commit a49b74f92a4e7d701d6f6cf63d207a8aff2e0f68.
`Module::getOrInsertGlobal` returns a `Constant*`, which is a super
class of `GlobalVariable`, but if the given type doesn't match an
existing declaration, it returns a bitcast of that global instead.
This causes UB when we pass that to `LLVMGetVisibility` which
unconditionally casts the opaque argument to a `GlobalValue*`.
Instead, we can do our own get-or-insert without worrying whether
existing types match exactly. It's not relevant when we're just trying
to get/set the linkage and visibility, and if types are needed we can
bitcast or error nicely from `rustc_codegen_llvm` instead.
The default diagnostic handler considers all remarks to be disabled by
default unless configured otherwise through LLVM internal flags:
`-pass-remarks`, `-pass-remarks-missed`, and `-pass-remarks-analysis`.
This behaviour makes `-Cremark` ineffective on its own.
Fix this by configuring a custom diagnostic handler that enables
optimization remarks based on the value of `-Cremark` option. With
`-Cremark=all` enabling all remarks.
In https://reviews.llvm.org/D71059 LLVM 11, the time trace profiler was
extended to support multiple threads.
`timeTraceProfilerInitialize` creates a thread local profiler instance.
When a thread finishes `timeTraceProfilerFinishThread` moves a thread
local instance into a global collection of instances. Finally when all
codegen work is complete `timeTraceProfilerWrite` writes data from the
current thread local instance and the instances in global collection
of instances.
Previously, the profiler was intialized on a single thread only. Since
this thread performs no code generation on its own, the resulting
profile was empty.
Update LLVM codegen to initialize & finish time trace profiler on each
code generation thread.
Add -Z no-unique-section-names to reduce ELF header bloat.
This change adds a new compiler flag that can help reduce the size of ELF binaries that contain many functions.
By default, when enabling function sections (which is the default for most targets), the LLVM backend will generate different section names for each function. For example, a function `func` would generate a section called `.text.func`. Normally this is fine because the linker will merge all those sections into a single one in the binary. However, starting with [LLVM 12](https://github.com/llvm/llvm-project/commit/ee5d1a04), the backend will also generate unique section names for exception handling, resulting in thousands of `.gcc_except_table.*` sections ending up in the final binary because some linkers like LLD don't currently merge or strip these EH sections (see discussion [here](https://reviews.llvm.org/D83655)). This can bloat the ELF headers and string table significantly in binaries that contain many functions.
The new option is analogous to Clang's `-fno-unique-section-names`, and instructs LLVM to generate the same `.text` and `.gcc_except_table` section for each function, resulting in a smaller final binary.
The motivation to add this new option was because we have a binary that ended up with so many ELF sections (over 65,000) that it broke some existing ELF tools, which couldn't handle so many sections.
Here's our old binary:
```
$ readelf --sections old.elf | head -1
There are 71746 section headers, starting at offset 0x2a246508:
$ readelf --sections old.elf | grep shstrtab
[71742] .shstrtab STRTAB 0000000000000000 2977204c ad44bb 00 0 0 1
```
That's an 11MB+ string table. Here's the new binary using this option:
```
$ readelf --sections new.elf | head -1
There are 43 section headers, starting at offset 0x29143ca8:
$ readelf --sections new.elf | grep shstrtab
[40] .shstrtab STRTAB 0000000000000000 29143acc 0001db 00 0 0 1
```
The whole binary size went down by over 20MB, which is quite significant.
This change adds a new compiler flag that can help reduce the size of
ELF binaries that contain many functions.
By default, when enabling function sections (which is the default for most
targets), the LLVM backend will generate different section names for each
function. For example, a function "func" would generate a section called
".text.func". Normally this is fine because the linker will merge all those
sections into a single one in the binary. However, starting with LLVM 12
(llvm/llvm-project@ee5d1a0), the backend will
also generate unique section names for exception handling, resulting in
thousands of ".gcc_except_table.*" sections ending up in the final binary
because some linkers don't currently merge or strip these EH sections.
This can bloat the ELF headers and string table significantly in
binaries that contain many functions.
The new option is analogous to Clang's -fno-unique-section-names, and
instructs LLVM to generate the same ".text" and ".gcc_except_table"
section for each function, resulting in smaller object files and
potentially a smaller final binary.
Implement `#[link_ordinal(n)]`
Allows the use of `#[link_ordinal(n)]` with `#[link(kind = "raw-dylib")]`, allowing Rust to link against DLLs that export symbols by ordinal rather than by name. As long as the ordinal matches, the name of the function in Rust is not required to match the name of the corresponding function in the exporting DLL.
Part of #58713.
Enable AutoFDO.
This largely involves implementing the options debug-info-for-profiling
and profile-sample-use and forwarding them on to LLVM.
AutoFDO can be used on x86-64 Linux like this:
rustc -O -Clink-arg='Wl,--no-rosegment' -Cdebug-info-for-profiling main.rs -o main
perf record -b ./main
create_llvm_prof --binary=main --out=code.prof
rustc -O -Cprofile-sample-use=code.prof main.rs -o main2
Now `main2` will have feedback directed optimization applied to it.
The create_llvm_prof tool can be obtained from this github repository:
https://github.com/google/autofdo
The option -Clink-arg='Wl,--no-rosegment' is necessary to avoid lld
putting an extra RO segment before the executable code, which would make
the binary silently incompatible with create_llvm_prof.
This largely involves implementing the options debug-info-for-profiling
and profile-sample-use and forwarding them on to LLVM.
AutoFDO can be used on x86-64 Linux like this:
rustc -O -Cdebug-info-for-profiling main.rs -o main
perf record -b ./main
create_llvm_prof --binary=main --out=code.prof
rustc -O -Cprofile-sample-use=code.prof main.rs -o main2
Now `main2` will have feedback directed optimization applied to it.
The create_llvm_prof tool can be obtained from this github repository:
https://github.com/google/autofdoFixes#64892.
No functional changes intended.
The LLVM commit
e463b69736
changed an argument of fatal_error_handler_t from std::string to char*.
This adapts RustWrapper accordingly.
thinLTOResolvePrevailingInModule became thinLTOFinalizeInModule and
gained the ability to propagate noRecurse and noUnwind function
attributes. I ran codegen tests with it both on and off, as the upstream
patch uses it in both modes, and the tests pass both ways. Given that,
it seemed reasonable to go ahead and let the propagation be enabled in
rustc, and see what happens. See https://reviews.llvm.org/D36850 for
more examples of how the new version of the function gets used.
Change ab41eef9aca3 in LLVM split MemorySanitizerPass into
MemorySanitizerPass for functions and ModuleMemorySanitizerPass for
modules. There's a related change for ThreadSanitizerPass, and in here
since we're using a ModulePassManager I only add the module flavor of
the pass on LLVM 14.
r? @nikic cc @nagisa
These were deleted in https://reviews.llvm.org/D108614, and in C++ I
definitely see the argument for their removal. I didn't try and
propagate the changes up into higher layers of rustc in this change
because my initial goal was to get rustc working against LLVM HEAD
promptly, but I'm happy to follow up with some refactoring to make the
API on the Rust side match the LLVM API more directly (though the way
the enum works in Rust makes the API less scary IMO).
r? @nagisa cc @nikic
The above-mentioned commit (part of the LLVM 14 development cycle)
removes a method that rustc uses somewhat extensively. We mostly switch
to lower-level methods that exist in all versions of LLVM we use, so no
new ifdef logic is required in most cases.
PassWrapper: adapt for LLVM 14 changes
These API changes appear to have all taken place in
https://reviews.llvm.org/D105007, which moved HWAddressSanitizerPass and
AddressSanitizerPass to only accept their options type as a ctor
argument instead of the sequence of bools etc. This required a couple of
parameter additions, which I made match the default prior to the
mentioned upstream LLVM change.
This patch restores rustc to building (though not quite passing all
tests, I've mailed other patches for those issues) against LLVM HEAD.
These API changes appear to have all taken place in
https://reviews.llvm.org/D105007, which moved HWAddressSanitizerPass and
AddressSanitizerPass to only accept their options type as a ctor
argument instead of the sequence of bools etc. This required a couple of
parameter additions, which I made match the default prior to the
mentioned upstream LLVM change.
This patch restores rustc to building (though not quite passing all
tests, I've mailed other patches for those issues) against LLVM HEAD.
PassWrapper: handle move of OptimizationLevel class out of PassBuilder
This is the first build break of the LLVM 14 cycle, and was caused by
https://reviews.llvm.org/D107025. Mercifully an easy fix.
Rather than relying on `getPointerElementType()` from LLVM function
pointers, we now pass the function type explicitly when building `call`
or `invoke` instructions.
This does not yet support #[link_name] attributes on functions, the #[link_ordinal]
attribute, #[link(kind = "raw-dylib")] on extern blocks in bin crates, or
stdcall functions on 32-bit x86.
This commit updates how rustc compiler metadata is stored in rlibs.
Previously metadata was stored as a raw file that has the same format as
`--emit metadata`. After this commit, however, the metadata is encoded
into a small object file which has one section which is the contents of
the metadata.
The motivation for this commit is to fix a common case where #83730
arises. The problem is that when rustc crates a `dylib` crate type it
needs to include entire rlib files into the dylib, so it passes
`--whole-archive` (or the equivalent) to the linker. The problem with
this, though, is that the linker will attempt to read all files in the
archive. If the metadata file were left as-is (today) then the linker
would generate an error saying it can't read the file. The previous
solution was to alter the rlib just before linking, creating a new
archive in a temporary directory which has the metadata file removed.
This problem from before this commit is now removed if the metadata file
is stored in an object file that the linker can read. The only caveat we
have to take care of is to ensure that the linker never actually
includes the contents of the object file into the final output. We apply
similar tricks as the `.llvmbc` bytecode sections to do this.
This involved changing the metadata loading code a bit, namely updating
some of the LLVM C APIs used to use non-deprecated ones and fiddling
with the lifetimes a bit to get everything to work out. Otherwise though
this isn't intended to be a functional change really, only that metadata
is stored differently in archives now.
This should end up fixing #83730 because by default dylibs will no
longer have their rlib dependencies "altered" meaning that
split-debuginfo will continue to have valid paths pointing at the
original rlibs. (note that we still "alter" rlibs if LTO is enabled to
remove Rust object files and we also "alter" for the #[link(cfg)]
feature, but that's rarely used).
Closes#83730
PassWrapper: update for LLVM change D102093
In https://reviews.llvm.org/D102093 lots of things stopped taking the
DebugLogging boolean parameter. Mercifully we appear to always set
DebugPassManager to false, so I don't think we're losing anything by not
passing this parameter.
In https://reviews.llvm.org/D102093 lots of things stopped taking the
DebugLogging boolean parameter. Mercifully we appear to always set
DebugPassManager to false, so I don't think we're losing anything by not
passing this parameter.
This lets me build against llvm/main as of March 23rd, 2021. I'm not
entirely sure this is _correct_, but it appears to be functionally
identical to what was done in LLVM: existing callsites of
setInlineAsmDiagnosticHandler were moved to SetDiagnosticHandler() on
the context object, which we already set up in both places that we
called setInlineAsmDiagnosticHandler().
`fast-math` implies things like functions not being able to accept as an
argument or return as a result, say, `inf` which made these functions
confusingly named or behaving incorrectly, depending on how you
interpret it. Since the time when these intrinsics have been implemented
the intrinsics user's (stdsimd) approach has changed significantly and
so now it is required that these intrinsics operate normally rather than
in "whatever" way.
Fixes#84268
This should have no real effect in most cases, as e.g. `hidden`
visibility already implies `dso_local` (or at least LLVM IR does not
preserve the `dso_local` setting if the item is already `hidden`), but
it should fix `-Crelocation-model=static` and improve codegen in
executables.
Note that this PR does not exhaustively port the logic in [clang]. Only
the obviously correct portion and what is necessary to fix a regression
from LLVM 12 that relates to `-Crelocation_model=static`.
Fixes#83335
[clang]: 3001d080c8/clang/lib/CodeGen/CodeGenModule.cpp (L945-L1039)
THis came up in the review of #83425: it's hard to imagine a use of
LLVM_VERSION_LE() or LLVM_VERSION_EQ() that's not asking for trouble
when a point release gets created, so let's just discard them to prevent
the issue.
As far as I can tell what we've been getting is llvm::MaybeAlign(), so
just use that for now. This is required sometime after
24539f1ef2471d07bd87f833cb0288fc0f251f4b.
This changed in 54fb3ca96e261f7107cb1b5778c34cb0e0808be6 - I'm not
entirely sure it's correct that we're leaving config empty, but the one
case in LLVM that looked similar did that.
This works around a design defect in the LLVM 12 pass builder
implementation. In LLVM 13, the PreLink ThinLTO pipeline properly
respects the OptimizerLastEPCallbacks.
The definition of this struct changes in LLVM 12 due to the addition
of branch coverage support. To avoid future mismatches, declare our
own struct and then convert between them.
This commit adds a new ABI to be selected via `extern
"C-cmse-nonsecure-call"` on function pointers in order for the compiler to
apply the corresponding cmse_nonsecure_call callsite attribute.
For Armv8-M targets supporting TrustZone-M, this will perform a
non-secure function call by saving, clearing and calling a non-secure
function pointer using the BLXNS instruction.
See the page on the unstable book for details.
Signed-off-by: Hugues de Valon <hugues.devalon@arm.com>
This commit modifies the FFI bindings to LLVM required for Split DWARF
support in rustc. In particular:
- `addPassesToEmitFile`'s wrapper, `LLVMRustWriteOutputFile` now takes
a `DwoPath` `const char*`. When disabled, `nullptr` should be provided
which will preserve existing behaviour. When enabled, the path to the
`.dwo` file should be provided.
- `createCompileUnit`'s wrapper, `LLVMRustDIBuilderCreateCompileUnit`
now has two additional arguments, for the `DWOId` and to enable
`SplitDebugInlining`. `DWOId` should always be zero.
- `createTargetMachine`'s wrapper, `LLVMRustCreateTargetMachine` has an
additional argument which should be provided the path to the `.dwo`
when enabled.
Signed-off-by: David Wood <david@davidtw.co>
* `rustc` should now compile under LLVM 9 or 10
* Compiler generates an error if `-Z instrument-coverage` is specified
but LLVM version is less than 11
* Coverage tests that require `-Z instrument-coverage` and run codegen
should be skipped if LLVM version is less than 11
Changes the coverage map injected into binaries compiled with
`-Zinstrument-coverage` to LLVM Coverage Mapping Format, Version 4 (from
Version 3). Note, binaries compiled with this version will require LLVM
tools from at least LLVM Version 11.
This commit grepped for LLVM_VERSION_GE, LLVM_VERSION_LT, get_major_version and
min-llvm-version and statically evaluated every expression possible
(and sensible) assuming that the LLVM version is >=9 now
Add support for SHA256 source file hashing
Adds support for `-Z src-hash-algorithm sha256`, which became available in LLVM 11.
Using an older version of LLVM will cause an error `invalid checksum kind` if the hash algorithm is set to sha256.
r? `@eddyb`
cc #70401 `@est31`
rustc_llvm: unwrap LLVMMetadataRef before casting
Directly casting the opaque pointer was [reported] to cause an
"incomplete type" error with GCC 9.3:
```
llvm-wrapper/RustWrapper.cpp:939:31: required from here
/usr/include/c++/9.3/type_traits:1301:12: error: invalid use of incomplete type 'struct LLVMOpaqueMetadata'
1301 | struct is_base_of
| ^~~~~~~~~~
In file included from [...]/rust/src/llvm-project/llvm/include/llvm-c/BitReader.h:23,
from llvm-wrapper/LLVMWrapper.h:1,
from llvm-wrapper/RustWrapper.cpp:1:
[...]/rust/src/llvm-project/llvm/include/llvm-c/Types.h:89:16: note: forward declaration of 'struct LLVMOpaqueMetadata'
89 | typedef struct LLVMOpaqueMetadata *LLVMMetadataRef;
| ^~~~~~~~~~~~~~~~~~
```
[reported]: https://zulip-archive.rust-lang.org/182449tcompilerhelp/12215halprustcllvmbuildfail.html#214915124
A simple `unwrap` fixes the issue.
r? `@eddyb`
Directly casting the opaque pointer was [reported] to cause an
"incomplete type" error with GCC 9.3:
```
llvm-wrapper/RustWrapper.cpp:939:31: required from here
/usr/include/c++/9.3/type_traits:1301:12: error: invalid use of incomplete type 'struct LLVMOpaqueMetadata'
1301 | struct is_base_of
| ^~~~~~~~~~
In file included from [...]/rust/src/llvm-project/llvm/include/llvm-c/BitReader.h:23,
from llvm-wrapper/LLVMWrapper.h:1,
from llvm-wrapper/RustWrapper.cpp:1:
[...]/rust/src/llvm-project/llvm/include/llvm-c/Types.h:89:16: note: forward declaration of 'struct LLVMOpaqueMetadata'
89 | typedef struct LLVMOpaqueMetadata *LLVMMetadataRef;
| ^~~~~~~~~~~~~~~~~~
```
[reported]: https://zulip-archive.rust-lang.org/182449tcompilerhelp/12215halprustcllvmbuildfail.html#214915124
A simple `unwrap` fixes the issue.
Use unwrapDIPtr because the Scope may be null.
I ran into an assertion when using debug information on Windows with LLVM assertions enabled.
It seems like we are using unwrap here (which in turn calls isa and requires the pointer to be non-null) but we expect the value to be null because that is what we are passing from rustc.
This change uses unwrapDIPtr which explicitly allows nullptr.
The FFI prototype for this method on the rust side has the `LLVMMetadataRef` parameter as `Scope: Option<&'a DIScope>`, and we always pass `None` when `msvc_like_names` is true.
Use llvm::computeLTOCacheKey to determine post-ThinLTO CGU reuse
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
Secure entry functions do not support if arguments are passed on the
stack. An "unsupported" diagnostic will be emitted by LLVM if that is
the case.
This commits adds support in Rust for that diagnostic so that an error
will be output if that is the case!
Signed-off-by: Hugues de Valon <hugues.devalon@arm.com>
During incremental ThinLTO compilation, we attempt to re-use the
optimized (post-ThinLTO) bitcode file for a module if it is 'safe' to do
so.
Up until now, 'safe' has meant that the set of modules that our current
modules imports from/exports to is unchanged from the previous
compilation session. See PR #67020 and PR #71131 for more details.
However, this turns out be insufficient to guarantee that it's safe
to reuse the post-LTO module (i.e. that optimizing the pre-LTO module
would produce the same result). When LLVM optimizes a module during
ThinLTO, it may look at other information from the 'module index', such
as whether a (non-imported!) global variable is used. If this
information changes between compilation runs, we may end up re-using an
optimized module that (for example) had dead-code elimination run on a
function that is now used by another module.
Fortunately, LLVM implements its own ThinLTO module cache, which is used
when ThinLTO is performed by a linker plugin (e.g. when clang is used to
compile a C proect). Using this cache directly would require extensive
refactoring of our code - but fortunately for us, LLVM provides a
function that does exactly what we need.
The function `llvm::computeLTOCacheKey` is used to compute a SHA-1 hash
from all data that might influence the result of ThinLTO on a module.
In addition to the module imports/exports that we manually track, it
also hashes information about global variables (e.g. their liveness)
which might be used during optimization. By using this function, we
shouldn't have to worry about new LLVM passes breaking our module re-use
behavior.
In LLVM, the output of this function forms part of the filename used to
store the post-ThinLTO module. To keep our current filename structure
intact, this PR just writes out the mapping 'CGU name -> Hash' to a
file. To determine if a post-LTO module should be reused, we compare
hashes from the previous session.
This should unblock PR #75199 - by sheer chance, it seems to have hit
this issue due to the particular CGU partitioning and optimization
decisions that end up getting made.
